Output chaining (fan-in → genText)

Run: npm run run:output-chaining
Pipeline: examples/pipeline-output-chaining.json

Output chaining — two webpage nodes fetch in parallel, fan in to genText (same node.type as arcpx-infra exports). Bedrock credentials are passed as GraphEngine({ llm }) — locally from .env.bedrock; in your app from code. See LLM configuration.

Fan-in with named outputTarget → {$ontario_gov} / {$canada_holidays} in prompt:

Node	node.type	Role
fetch_example	webpage	Ontario.ca
fetch_wikipedia	webpage	Statutory holidays
genText_we_are_canada	genText	Bedrock synthesis (ArcPX-compatible JSON)

ArcPX export compatibility

Use the same fields ArcPX puts in the graph:

• type: "genText"
• data.isPassThrough: true
• data.nodeData: string prompt — leading System Prompt: block plus {$ontario_gov}, {$canada_holidays} (names match upstream outputTarget without $); instructs the model to print sources (title + URL) in a Sources section
• data.settings: modelId, region, optional outputPath
• Upstream webpage nodes set outputTarget (e.g. "$ontario_gov") so genText can reference each fetch by name

Export JSON from ArcPX → run with:

cp .env.bedrock.template .env.bedrock
npm run run:output-chaining
# or: node examples/run-with-llm.mjs ./your-exported-pipeline.json

Programmatic (npm package)

import {
  GraphEngine,
  loadFlowFromFile,
  createBuiltinNodeExecutor,
} from "arc-dag";
import { createNodeExecutor } from "./examples/lib/handlers/index.mjs"; // + webpage

const flow = await loadFlowFromFile("./examples/pipeline-output-chaining.json");

const engine = new GraphEngine({
  flow,
  llm: {
    provider: "bedrock",
    apiKey: process.env.BEDROCK_API_KEY!,
    modelId: "us.anthropic.claude-sonnet-4-6",
    region: "us-east-1",
  },
  nodeExecutor: createNodeExecutor(), // package builtin + webpage handler
});

await engine.run();

Fan-in

genText receives data.sourceData as an array of both fetch results. The handler builds the prompt (substituting {$ontario_gov}, {$canada_holidays}) and calls Bedrock Converse using llm settings on the engine.

Output

• LLM response (string or { text, writtenTo } when settings.outputPath is set)
• File: examples/output/we-are-canada.md (rendered model output)

Handlers

• web-scraper/ — webpage
• llm-bedrock/ — genText + LLM → Bedrock

LLM configuration · Node handlers · BYO LLM